SGER: Gradual Compilers

点击翻译按钮获取中文摘要

CCR-9987458Abstract:With the increasing demands of users for ever more complex features intheir systems, interfaces to these systems necessarily also becomecomplex. At some point, the interfaces approximate a programminglanguage, and the system can reasonably use a compiler to efficientlyhandle the users' directions. We thus expect that compiler-writingtechnology will increasingly be demanded of software engineeringprofessionals. Unfortunately, compilers are often large, mysteriousprograms, and compiler-writing somewhat of a black art. We are developing a new approach to the development and exposition ofcompilers for mostly functional languages. Courses on compilers aregenerally centered around an extensive project through which acompiler is developed. Some attempt is often made to divide thecompiler into phases. Traditionally, however, the early phases merelycover scanning and parsing and most of the real work is left as asingle task. This makes compiler courses either difficult, ifstudents are left to deal with this problem on their own, orcontrived, if they are given substantial amounts of help from theirinstructor. The primary advantage of our approach is that studentscan build a compiler in incremental stages and understand how eachstage contributes to the project. Thus, at any point in time, thestudents are only concerned with a single aspect of the targetarchitecture. In this way, they gain an understanding of how thesource and target architectures relate to each other and no longerview them as totally disparate worlds. An additional advantage of ourapproach is that instructions at different levels can be interleaved,so that transformations can be written and debugged for one languageform at a time. All of this will make compiler-writing far lessmysterious and make compiler courses accessible to many more students.We are submitting this request under Small Grants for ExploratoryResearch because we wish to pursue this research immediately. Thisnew approach, if further developed, could greatly affect IndianaUniversity's undergraduate compiler course this spring. We will alsomake the results of our efforts available to the larger educationalcommunity through submissions to education-oriented conferences and/orthe publication of relevant texts.Although our primary motivation at this point is educational, thebenefits of our approach are likely to range much further. Theyshould include the ability to reuse components in compilers with adifferent source or target language, to the point where writing acompiler involves plotting a course through a variety of intermediatelanguages or programming styles or architectures. Then, a change toany one architecture will have minimal impact on the family ofcompilers. A second advantage is that optimizations can be addedindividually, at the highest appropriate level. Finally, support forthe interleaving of instructions at different levels should make thedevelopment of compilers much less expensive and error-prone. Itshould also help to improve intuitions and understanding of issues of just-in-time compililation that have become pressing withthe increased popularity of the Java programming language.